@absolutejs/voice-speechmatics

Speechmatics real-time speech-to-text adapter for @absolutejs/voice.

Speaks Speechmatics' v2 WebSocket protocol directly: connect → StartRecognition → binary audio frames → AddPartialTranscript / AddTranscript / EndOfTranscript → EndOfStream. No Speechmatics SDK dependency.

Install

bun add @absolutejs/voice-speechmatics

@absolutejs/voice is a runtime dependency.

Use

import { voice } from "@absolutejs/voice";
import { speechmatics } from "@absolutejs/voice-speechmatics";

const app = voice({
  stt: speechmatics({
    apiKey: process.env.SPEECHMATICS_API_KEY!,
    region: "eu2",                       // 'eu' | 'eu2' | 'usa'
    language: "en",
    operatingPoint: "enhanced",          // or 'standard' for lower latency
    enablePartials: true,
    diarization: "speaker",              // optional
  }),
  // ... tts + other options ...
});

For pre-issued JWT tokens (e.g., short-lived browser sessions):

speechmatics({ region: "eu2", jwt: shortLivedJwt, language: "en" });

Options

| Option | Required | Default | Notes | | --- | --- | --- | --- | | apiKey / jwt | one of | — | API key or pre-issued JWT; passed as ?jwt=<token> on the WebSocket URL per the Speechmatics protocol. | | region | no | eu2 | eu, eu2, usa, or any region prefix that resolves to wss://{region}.rt.speechmatics.com. | | baseUrl | no | wss://{region}.rt.speechmatics.com | Override for private/staging endpoints. | | language | no | en | BCP-47 code. Overridden when STTAdapterOpenOptions.languageStrategy fixes a language. | | operatingPoint | no | enhanced | enhanced (higher quality) or standard (lower latency). | | enablePartials | no | true | Stream partial hypotheses. | | diarization | no | — | speaker, channel, channel_and_speaker, or none. | | maxDelay, speakerChangeSensitivity, punctuationOverrides | no | — | Forwarded to transcription_config. | | connectTimeoutMs | no | 8000 | Time to wait for the RecognitionStarted ACK. | | webSocket.factory | no | new WebSocket(url) | Inject a fake socket for tests. |

Notes

• Audio sent before the RecognitionStarted ACK is buffered and flushed once recognition starts.
• AddPartialTranscript → partial event, AddTranscript → final event, EndOfTranscript → endOfTurn.
• Punctuation tokens are joined to the preceding word (Hello, world, . → Hello world.).
• Average per-token confidence is lifted onto the transcript.
• session.close(reason) sends EndOfStream with the last seq_no and then closes the socket cleanly.
• Supported encodings on STTAdapterOpenOptions.format.encoding: pcm_s16le, pcm_f32le, mulaw / pcm_mulaw.